Re: [PATCH] tracing: Do not synchronize freeing of trigger filter on boot up

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

On 2022-12-14 15:03, Paul E. McKenney wrote:
> On Wed, Dec 14, 2022 at 08:49:54AM +0900, Masami Hiramatsu wrote:
>> On Tue, 13 Dec 2022 17:24:29 -0500
>> Steven Rostedt <rostedt@goodmis.org> wrote:
>>
>>> From: "Steven Rostedt (Google)" <rostedt@goodmis.org>
>>>
>>> If a trigger filter on the kernel command line fails to apply (due to
>>> syntax error), it will be freed. The freeing will call
>>> tracepoint_synchronize_unregister(), but this is not needed during early
>>> boot up, and will even trigger a lockdep splat.
>>>
>>> Avoid calling the synchronization function when system_state is
>>> SYSTEM_BOOTING.
>>
>> Shouldn't this be done inside tracepoint_synchronize_unregister()?
>> Then, it will prevent similar warnings if we expand boot time feature.
> 
> How about the following wide-spectrum fix within RCU_LOCKDEP_WARN()?
> Just in case there are ever additional issues of this sort?

Hi Paul,

Your approach makes sense. Thanks for looking into this.

Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

> 
> 							Thanx, Paul
> 
> ------------------------------------------------------------------------
> 
> commit d493ffca2df6c1963bd1d7b8f8c652a172f095ae
> Author: Paul E. McKenney <paulmck@kernel.org>
> Date:   Wed Dec 14 11:41:44 2022 -0800
> 
>      rcu: Make RCU_LOCKDEP_WARN() avoid early lockdep checks
>      
>      Currently, RCU_LOCKDEP_WARN() checks the condition before checking
>      to see if lockdep is still enabled.  This is necessary to avoid the
>      false-positive splats fixed by commit 3066820034b5dd ("rcu: Reject
>      RCU_LOCKDEP_WARN() false positives").  However, the current state can
>      result in false-positive splats during early boot before lockdep is fully
>      initialized.  This commit therefore checks debug_lockdep_rcu_enabled()
>      both before and after checking the condition, thus avoiding both sets
>      of false-positive error reports.
>      
>      Reported-by: Steven Rostedt <rostedt@goodmis.org>
>      Reported-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
>      Reported-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
>      Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
>      Cc: Boqun Feng <boqun.feng@gmail.com>
>      Cc: Matthew Wilcox <willy@infradead.org>
>      Cc: Thomas Gleixner <tglx@linutronix.de>
> 
> diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> index aa1b1c3546d7a..1aec1d53b0c91 100644
> --- a/include/linux/rcupdate.h
> +++ b/include/linux/rcupdate.h
> @@ -364,11 +364,18 @@ static inline int debug_lockdep_rcu_enabled(void)
>    * RCU_LOCKDEP_WARN - emit lockdep splat if specified condition is met
>    * @c: condition to check
>    * @s: informative message
> + *
> + * This checks debug_lockdep_rcu_enabled() before checking (c) to
> + * prevent early boot splats due to lockdep not yet being initialized,
> + * and rechecks it after checking (c) to prevent false-positive splats
> + * due to races with lockdep being disabled.  See commit 3066820034b5dd
> + * ("rcu: Reject RCU_LOCKDEP_WARN() false positives") for more detail.
>    */
>   #define RCU_LOCKDEP_WARN(c, s)						\
>   	do {								\
>   		static bool __section(".data.unlikely") __warned;	\
> -		if ((c) && debug_lockdep_rcu_enabled() && !__warned) {	\
> +		if (debug_lockdep_rcu_enabled() && (c) &&		\
> +		    debug_lockdep_rcu_enabled() && !__warned) {		\
>   			__warned = true;				\
>   			lockdep_rcu_suspicious(__FILE__, __LINE__, s);	\
>   		}							\

-- 
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com